1 paper
Marah Abdin, Jyoti Aneja, Hany Awadalla +126
We introduce phi-3-mini, a 3.8 billion parameter language model trained on 3.3 trillion tokens, whose overall performance, as measured by both academic benchmarks and internal test…